A control method for controlling a speed of a movable machine element of a numerically controlled industrial processing machine with jerk limitation is described. The processing machine can be, for example, a machine tool, a production machine and/or a robot. A travel path of the movable machine element is first decomposed into a plurality of sequentially arranged, interpolatable travel sections. Jerk profiles are then determined for the interpolatable travel sections, which are modified with a parameter-dependent shape function. The modified jerk profiles provide a filter action that can be represented by bandpass filters having blocking frequencies. The blocking frequencies of the bandpass filters are selected so as to substantially correspond to characteristic frequencies of the machine element. The control method significantly improves the machine dynamics without sacrificing accuracy.

Patent
   7208901
Priority
Apr 04 2003
Filed
Apr 02 2004
Issued
Apr 24 2007
Expiry
Jun 03 2025
Extension
427 days
Assg.orig
Entity
Large
6
10
EXPIRED
1. A control method for controlling a speed of a movable machine element of a numerically controlled industrial processing machine with jerk limitation, comprising the steps of:
decomposing a travel path of the movable machine element into a plurality of sequentially arranged, interpolatable travel sections;
determining jerk profiles for the interpolatable travel sections; and
modifying the jerk profiles with a parameter-dependent shape function obtained by adding a rectangular shape function and a Dirac shape function, with the modified jerk profile in a corresponding travel section providing a filter action represented by a bandpass filter having at least one blocking frequency,
wherein the rectangular shape function and the Dirac shape function are each adjusted with a respective parameter, so that the at least one blocking frequency of the bandpass filter corresponds substantially to a characteristic frequency of the machine element.
11. A control method for controlling a speed of a movable machine element of a numerically controlled industrial processing machine with jerk limitation, comprising the steps of:
decomposing a travel path of the movable machine element into a plurality of sequentially arranged, interpolatable travel sections;
determining jerk profiles for the interpolatable travel sections; and
modifying the jerk profiles with a parameter-dependent shape function obtained by adding a rectangular shape function and a cos2 or sin2 shape function, with the modified jerk profile in a corresponding travel section providing a filter action represented by a bandpass filter having at least one blocking frequency,
wherein the rectangular shape function and the cos2 or sin2 shape function are each adjusted with a respective parameter, so that the at least one blocking frequency of the bandpass filter corresponds substantially to a characteristic frequency of the machine element.
2. The control method of claim 1, wherein the parameter-dependent shape function is defined by a plurality of parameters.
3. The control method of claim 2, wherein the parameter-dependent shape function is formed by adding a plurality of rectangular shape functions and Dirac shape functions, with each of the plurality of shape functions being weighted with a corresponding one of the plurality of parameters.
4. The control method of claim 3, and further adding weighted shape functions selected from the group consisting of cos2 or sin2 shape function, trapezoidal shape function, and triangular shape function.
5. The control method of claim 2, wherein the plurality of parameters are selected so as to define a plurality of blocking frequencies.
6. The control method of claim 1, wherein the parameter-dependent shape function is defined by a single parameter.
7. The control method of claim 6, wherein the parameter-dependent shape function is obtained by adding the rectangular shape function and the Dirac shape function, wherein weighting between the rectangular shape function and Dirac shape function is performed by the single parameter.
8. The control method of claim 1, wherein the shape of the jerk profile is modified by the parameter-dependent shape function so that an area of the jerk profile after the modification is identical to an area of the jerk profile before the modification, computed over an identical time interval.
9. The control method of claim 1, and further comprising the steps of before executing a defined travel path of the movable machine element, determining rectangular jerk profiles optimized for the corresponding interpolatable travel sections, such that the blocking frequencies of the bandpass filters coincide essentially with the characteristic frequencies of the machine element, and then further optimizing the previously optimized rectangular jerk profile with the parameter-dependent shape function.
10. The control method of claim 1, wherein the industrial processing machine comprises a machine selected from the group consisting of machine tool, production machine and robot.
12. The control method of claim 11, wherein the parameter-dependent shape function is defined by a single parameter.
13. The control method of claim 12, wherein the parameter-dependent shape function is obtained by adding the rectangular shape function and cos2 or sin2 shape function, with weighting between the rectangular and cos2 or sin2 shape functions performed by the single parameter.
14. The control method of claim 11, wherein the parameter-dependent shape function is defined by a plurality of parameters.
15. The control method of claim 14, wherein the parameter-dependent shape function is formed by adding a plurality of rectangular shape functions and cos2 or sin2 shape functions, with each of the plurality of shape functions being weighted with a corresponding one of the plurality of parameters.
16. The control method of claim 15, and further adding weighted shape functions selected from the group consisting of Dirac shape function, trapezoidal shape function, and triangular shape function.
17. The control method of claim 14, wherein the plurality of parameters are selected so as to define a plurality of blocking frequencies.
18. The control method of claim 11, wherein the industrial processing machine comprises a machine selected from the group consisting of machine tool, production machine and robot.

This application claims the priority of German Patent Application, Serial No. 103 15 525.2, filed Apr. 4, 2003, pursuant to 35 U.S.C. 119(a)–(d), the disclosure of which is incorporated herein by reference.

The present invention relates to a control method for jerk-limited speed control of a movable machine element of a numerically controlled industrial processing machine, such as a machine tool, production machine or a robot. In the context of this application, the term “production machine” is used in a generic sense and also includes robots which generally follow the concepts outlined here.

A mechanical system that is capable of performing oscillations (e.g., along the axes of the machine tool) is typically characterized by at least one characteristic frequency which is excited by a movement or displacement and which can be observed in the actual position value. Excitations of the various mechanisms of a machine should therefore be eliminated.

Conventional numerical controllers that control the speed of, e.g., a machine tool, production machine or a robot, typically include a jerk limitation. The goal is to reduce the load on the various axes of the machine without adversely affecting the program processing time.

A jerk limitation can slow down the buildup of the acceleration of a machine movement so as to smooth the desired value and to move the machine mechanism with the smallest possible oscillation excitation.

However, the smoothing effect of a jerk limitation depends strongly on the desired curve for the setpoint value. Measurements and theoretical studies have shown that higher-frequency acceleration and braking processes have to be performed with a lower dynamic range so as keep the excitations of the oscillatable mechanism small. This dependence applies to both short-term positioning actions as well as, for example, for controlling the entire path of the machine.

Currently, this problem is addressed by adjusting the jerk limit value and the acceleration limit value to a low value, so that even high-frequency changes in the path speed do not significantly excite the oscillations. However, setting the dynamic values very low often prevents a higher path speed and hence also a shorter program processing time, which would be otherwise desirable.

A known method for producing a speed profile that protects the machine includes a jerk limitation. In phase 1, the machine or machine element is moved with the highest acceptable acceleration. In phase 2, the acceleration remains constant so that the speed increases linearly. In phase 3, the acceleration is then decreased. In the following, the term acceleration is meant to also include a possible negative acceleration.

Accordingly, a maximum allowable path speed is reached at the end of the phase 3, which is then used for moving the machine or machine elements during phase 4. The velocities now decrease in an analog fashion during additional phases 5 to 7, so that the path speed becomes zero when a desired position is reached. For this purpose, a negative acceleration is generated in phase 5, which is kept constant in phase 6 and again reduced to zero in phase 7. The slope of the acceleration in phases 1, 3, 5, and 7 is critical for the jerk limitation of the machine. The speed curve in these phases can be described by a polynomial, whereas in the other phases the speed is linear or constant.

As discussed above, the time dependence of the jerk r(t) is composed of constant rectangular basic shapes. FIG. 1 shows the time dependence of the jerk r(t) during the time intervals tr, ta, and tv, which represent the time intervals of the jerk r(t) for an exemplary motion path of a movable machine element. The jerk r(t) is either equal to zero or is equal to its permissible maximum value r0 with a positive or negative sign: r(t) with r(t)ε{0,+r0,−r0}.

The time-dependent curve of the jerk r(t) for a movement or a displacement of the movable machine element will in the following be referred to as a jerk profile r(t).

A control method for jerk-limited speed control of a movable machine element is known from the German patent application DE 102 00 680.6. This application suggests that the transient behavior of the movable machine elements can be improved by changing or adapting the jerk profile with the help of a sin2 shape function.

The aforementioned problem is that acceleration and braking processes with a high dynamic range along the travel path x of the machine element can still cause mechanical oscillations in a certain frequency range. The dynamic properties of these processes should therefore be adapted to the characteristic properties of the machine.

It would therefore be desirable and advantageous to provide a control method for controlling a speed of a movable machine element of a numerically controlled industrial processing machine, which obviates prior art shortcomings and specifically eliminates the excitation of characteristic resonance frequencies of a machine or a movable machine element.

The control method according to the present invention describes a new approach whereby, unlike with the conventional method, not a specific shape function is selected, but an optimal shape function is computed based on the dominant characteristic frequencies of the machine. In this way, the dynamic properties of the machine can be significantly improved over the conventionally attained dynamic properties of the machine without sacrificing accuracy.

According to one aspect of the invention, a control method for controlling a speed of a movable machine element of a numerically controlled industrial processing machine with jerk limitation is disclosed. The control method includes the steps of decomposing a travel path of the movable machine element into a plurality of sequentially arranged, interpolatable travel sections, determining jerk profiles for the interpolatable travel sections, and modifying the jerk profiles with a parameter-dependent shape function. The modified jerk profiles in a corresponding travel section provide a filter action represented by bandpass filters having blocking frequencies. The blocking frequencies of the bandpass filters correspond substantially to the characteristic frequencies of the machine element.

The excitation of resonances can be prevented by intentionally adapting the jerk curve with a parameter-dependent shape function. Not only can the dynamic limit values be significantly increased in this way, but the program processing time can also be reduced considerably.

A control method according to the invention has, i.a., the following advantages:

This produces much softer movements in machine tools, production machines and robots, a higher accuracy, e.g., for surface machining, and an improved quality. It also increases the machining speed, a higher parts throughput and a higher efficiency without sacrificing quality.

According to another advantageous feature of the invention, the parameter-dependent shape function can be defined by at least one parameter αi. In this way, the frequencies of the bandpass filters can be easily adapted to the characteristic frequencies of the machine element and/or the machine or a selected subset of the machine element and/or the machine.

According to another feature of the invention, the parameter-dependent shape function can be defined by a single parameter α. Since only a single parameter is provided for adapting the shape function, the frequency of a bandpass filter can be particularly easily adapted to the characteristic frequencies of the machine element and/or the machine.

The shape function and the resulting jerk profile can have a particularly advantageous shape when the frequency of the bandpass filter is adapted to the characteristic frequency of the machine element and/or the machine by adding a rectangular shape function and a cos2 or sin2 shape function and/or alternatively by adding a rectangular shape function and a Dirac shape function.

A loss of path length can be prevented by modifying the shape of the jerk profile with the parameter-dependent shape function so that the area of the jerk profile after the modification is identical to the area of the jerk profile before the modification, when computed over an identical time interval.

According to another advantageous feature of the invention, a rectangular jerk profile that is optimized for the corresponding interpolatable travel sections can be determined before having the movable machine element execute a defined travel path. The rectangular jerk profile is selected so that the blocking frequencies of the bandpass filters coincide essentially with the characteristic frequencies of the machine element. The optimized rectangular jerk profile is then further modified with the parameter-dependent shape function. This approach includes exhausts all possibilities for optimization and produces an optimized global result.

Other features and advantages of the present invention will be more readily apparent upon reading the following description of currently preferred exemplified embodiments of the invention with reference to the accompanying drawing, in which:

FIG. 1 is a graphical illustration of a rectangular jerk profile of a conventional jerk limitation;

FIG. 2 is a block diagram of a system capable of performing oscillations, with a jerk-limited desired position value and an oscillating actual position value;

FIG. 3 is a graphical illustration of an amplitude-frequency curve of the transfer function A(f) of a jerk-equivalent filter;

FIG. 4 is a graphical illustration of three exemplary shape functions;

FIG. 5 is a graphical illustration of three different jerk profiles;

FIG. 6 is a graphical illustration of an amplitude-frequency curve of transfer functions of various jerk-equivalent filters;

FIG. 7 is a graphical illustration of three transient responses; and

FIG. 8 is a graphical illustration of an amplitude-frequency curve of the corresponding shape functions for the responses of FIG. 7.

Throughout all the Figures, same or corresponding elements are generally indicated by same reference numerals. These depicted embodiments are to be understood as illustrative of the invention and not as limiting in any way. It should also be understood that the drawings are not necessarily to scale and that the embodiments are sometimes illustrated by graphic symbols, phantom lines, diagrammatic representations and fragmentary views. In certain instances, details which are not necessary for an understanding of the present invention or which render other details difficult to perceive may have been omitted.

Turning now to the drawing, and in particular to FIG. 1, there is shown an exemplary rectangular jerk profile r(t) as a function of time t. The term “jerk” herein refers to the time derivative of the acceleration a(t). The jerk profile r(t) is characterized by its maximum jerk height r0 as well as by the various times tr, ta and tv that describe the motion path of the movable machine element. In order to ensure a very stiff transient response of an actual position value xist(t) of a movable machine element after the machine element has moved, conventional control methods employ a so-called jerk limitation, whereby the maximum jerk is limited to a value of, e.g., r0. As described above, an improved jerk limitation can be achieved by modeling and adapting the jerk curve with a sin2 shape function.

Turning now to FIG. 2, an oscillatable system M with a jerk-limited desired position value xsoll(t) and an oscillating actual position value xist(t) is shown to systematically describe the inadequacy of a jerk limitation. The frequency dependence of the amplitude B(f) of the transfer function A(f) of the jerk-equivalent filter R is computed. The frequency dependence of the amplitude B(f) of the transfer function A(f) of the jerk-equivalent filter R can be determined as a function of the limit values (e.g., the maximum speed v0, the maximal acceleration a0, the maximum jerk R0) and the traveled path or the movement along the motion path x. The amplitude-frequency curve of the transfer function A(f) describes the excitation of the oscillatable system M over the entire frequency range.

The filter effect of the jerk limitation was systematically analyzed and optimized within the context of the invention. With the limit values r0, a0 and v0 preset by the controller, a desired output position value xsoll(t) is obtained for an arbitrarily set input position value usoll(t). The desired output position value xsoll(t) is supplied, for example, to a controller that controls the movement or the displacement of a machine element.

The exemplary amplitude-frequency curve B(f) of the transfer function A(f) for a jerk-equivalent filter R is obtained by dividing the Fourier-transformed desired position value by the Fourier-transform of the desired input position value usoll(t) and then computing the absolute value. The filter effect of the jerk limitation can hereby be determined for any mechanical system. For computing the amplitude-frequency curve B(f) of the transfer function A(f), according to FIG. 3, of the jerk-equivalent filter R, the jerk profile r(t) shown in FIG. 1 was first successively integrated three times to determine the desired position value xsoll(t).

The systematic approach of the invention for minimizing of the excitation of oscillations can be best understood from an analysis of the jerk limitation with the help of the jerk-equivalent filter R. This approach, which will be described below, is much more sophisticated than the commonly used jerk limitation and even the jerk limitation employing a sin2 shape function. Importantly, the excitation of oscillations of the movable machine elements, such as the tools of the machine, can be decreased by a parameter-dependent change of the shape of the jerk profile r(t). A so-called shape function F(t) is introduced and used to generate an optimized jerk profile r(t) as depicted in FIG. 5.

The invention describes a control method which, unlike the afore-described methods, does not preselect a special shape function F(t), but instead computes an optimal shape function F(t) based on the dominant characteristic frequencies of the machine mechanism. In this way, the dynamical characteristic of the machine can be significantly enhanced while still maintaining the required machine precision.

It can be mathematically shown that the transfer function A(f) of a jerk-equivalent filter R can be expressed by the equation
A(f)=S·[ℑ(F(t/tr))]·[sitafsi(π(ta+tv)f)]·e−jπf(tr+2ta+ty)  (1)
wherein

si ( x ) = sin ( x ) x ,

The factor S is the so-called multiplication factor of the shape function F(t/tr) and serves to adapt the amplitude of the shape function to the height of the jerk profile (normalization of the shape function F(t/tr)). The times tr, ta and tv are defined by the boundary conditions (maximum value of the speed v0, maximum acceleration a0, and maximum jerk r0) as well as by the displacement x. The object here is to change or adapt to the jerk profile r(t) and thereby modify the frequency-dependence of the amplitude B(f) of the transfer function A(f), so that the frequencies of the bandpass filters of the amplitude curve B(f) of the transfer function A(f) of the jerk-equivalent filter R coincide essentially with the characteristic frequencies of the machine elements and/or the machine or with a subset of such characteristic frequencies. In this way, the highest possible damping in the corresponding frequency ranges of the bandpass filter is obtained.

The afore-described approach should not impair the dynamic characteristics, i.e., the travel time should remain unchanged while also not diminishing the accuracy. The frequency-dependence of the amplitude curve B(f) of the transfer function A(f) of the jerk-equivalent filter R can be adapted by the shape function F(t/tr), as is evident from equation (1). A shape function F(t/tr) which has a filtering effect to prevent excitation of these frequencies, can be computed by taking into account the relevant characteristic frequencies of the oscillatable mechanical system. The term “bandpass filters” herein refers to the frequency ranges of the amplitude-frequency curve B(f) of the transfer function A(f) of the jerk-equivalent filter R where the amplitude-frequency curve B(f) of the transfer function A(f) of the jerk-equivalent filter R of FIG. 3 has peaks with a high damping. This is indicated in FIG. 3 by the notches in the frequency curve.

The shape function F(t) can be adapted to the machine characteristics by defining a so-called parameter-dependent shape function F(t, α), which has a shape that is defined by a single parameter α. It will be understood by those skilled in the art that other shape functions can also be selected that depend, for example, on several parameters.

The use of a parameter-dependent shape function Fs(t, α), which is defined by adding a rectangular shape function F1(t, α) and a cos2 shape function F2(t), will now be described. The parameter α determines the weight of the contributions from the two shape functions F1(t) and F2(t).

The rectangular shape function F1(t) is defined by equation (2). The rectangular shape function F1(t) is shown in FIG. 4 as curve 1d.

F 1 ( t ) = - t r 2 t r 2 ( t t r ) ( 2 )
wherein

a b ( x ) = 1
in the interval between a and b, and

a b ( x ) = 0
elsewhere.

The Fourier transform of the shape function F1(t) is defined by equation (3):

( F 1 ( t ) ) = sin ( π f t r ) π f t r ( 3 )

The time dependence of the cos2 shape function F2(t) is defined by equation (4) and is shown in FIG. 4 as curve 2d.

F 2 ( t ) = 2 · - t r 2 t r 2 ( t t r ) · cos 2 ( π t / t r ) ( 4 )

The Fourier transform of the shape function F2(t) is defined by equation (5):

( F 2 ( t ) ) = sin ( π f t r ) π f t r · 1 1 - ( f t r ) 2 ( 5 )

The two shape functions F1(t) and F2(t) are then added and weighted with a single parameter α, whereby the shape function F1(t) is weighted by multiplication with the parameter α, and the shape function F2(t) is weighted by multiplication with the parameter (1-α). The result is a parameter-dependent shape function Fs(t, α) according to equation (6):

F s ( t , α ) = α · - t r 2 t r 2 ( t t r ) + 2 ( 1 - α ) · - t r 2 t r 2 ( t t r ) · cos 2 ( π t / t r ) ( 6 )

The time dependence of the parameter-dependent shape function Fs(t) is shown in FIG. 4 as curve 3d.

The Fourier transform of the parameter-dependent shape function Fs(t, α) is defined by equation (7):

( F s ( t , α ) ) = sin ( π f t r ) π f t r · 1 - α ( f t r ) 2 1 - ( f t r ) 2 ( 7 )

The parameter-dependent shape function Fs(t, α) is obtained by adding the two shape functions F1(t) and F2(t) after weighting with the parameter α, as described above. An additional zero value in the frequency-space amplitude of the Fourier transform of the parameter-dependent shape function Fs(t, α), that depends on the parameter α and the jerk time tr, is obtained at the frequency

f 0 = 1 α · t r ( 8 )

The jerk time tr is typically defined by the motion characteristic and the travel path and cannot be arbitrarily changed, As a result, the jerk time tr in the iterations (6) to (8) is generally predefined and can no longer be viewed as a parameter that can be freely selected by the user.

The amplitude of the Fourier transform of the shape function Fs(t, α) depends on the parameter α in such a way that the zero value f0 can be moved across the entire frequency range by varying α. The parameter α therefore represents an additional degree of freedom.

The bandpass filters shown in FIG. 3 can be modified by changing the parameter α in such a way that the frequencies of the bandpass filters coincide essentially with the characteristic frequencies of the machine element and/or of the machine or with a subset of such characteristic frequencies. The bandpass filters indicate those frequency ranges where the amplitude of the frequency curve B(f) of the transfer function A(f) is highly damped (notch). The parameter α can be computed for each travel profile so that the oscillatable mechanism of the machine or of the movable machine element is not excited for a predefined frequency between 1/tr and infinity. The resonance frequencies of the mechanism can then be suppressed.

For α=0, the frequency f0 is infinite, which corresponds to the cos2 shape function F2(t). For α=1, the frequency f0 is exactly 1/tr, corresponding to the rectangular shape function F2(t).

In general, the entire frequency range can be covered by combining the two shape functions with in arbitrarily selected weighting factor α. However, a useful range for the weighting factor α is between 0 and 1. Selecting the weighting factor α between 0 and 1 has the following advantages:

FIG. 5 shows the jerk profiles 2a and 3a, which are modified by the shape functions, as well as the original unchanged rectangular jerk profile 1a. The cos2-shaped jerk profile 2a was modified with the cos2 shape function 2d depicted in FIG. 4 according to the equation (4). The parameter-dependent jerk profile 3a is obtained by modifying the rectangular jerk profile 1a with the parameter-dependent shape function Fs(t, α ) according to equation (6) which is visualized in FIG. 4 as curve 3d. In the exemplary embodiment, the parameter α was set to α=0.345. It should be noted that the shape of a jerk profile r(t) can be changed with the parameter-dependent shape function Fs(t, α ) in such a way that the area of the jerk profile r(t) before the shape change is identical to the area of the jerk profile r(t) after the shape change at the same time. This prevents a loss in the length of the travel path. For this reason, the shape function F2(t) is multiplied by a factor of 2.

FIG. 6 shows the frequency-dependent amplitude of the transfer function of the resulting jerk-equivalent filter obtained from the corresponding jerk profiles 1a, 2a, and 3a according to FIG. 5. The rectangular jerk profile 1a has a frequency-dependent amplitude curve 1b, the cos2-shaped jerk profile 2a has a frequency-dependent amplitude curve 2b, and the parameter-dependent jerk profile 3a has a frequency-dependent amplitude curve 3b.

In the illustrated exemplary embodiment, the resonance frequency of the mechanism of the machine is 24 Hz. As seen in FIG. 6, the frequency-dependent amplitude curve of the jerk profile 3a of FIG. 5 that was modified with the parameter-dependent shape function Fs(t, α) has a significantly higher damping at 24 Hz than the frequency-dependent amplitude curves of the jerk profiles 1a and 2a. The jerk profile 1a corresponds to a rectangular jerk profile, whereas the jerk profile 1b corresponds to a jerk profile that has been modified with the cos2 shape function F2(t). Accordingly, the machine components at the resonance frequency of 24 Hz are excited to a lesser degree when using a parameter-dependent jerk profile than with a cos2-shaped jerk profile or a purely rectangular jerk profile. This results in a faster and superior transient response characteristic from the actual position value to the desired position value.

FIG. 7 shows a graphic representation of the effect of modifying the jerk profiles 1a, 2a, and 3a according to the embodiment depicted in FIGS. 5 and 6. A movable machine element, e.g. a tool, is hereby moved from the position 0 to the desired initial position usoll(t)=40 mm. The corresponding actual position value xist(t) of the movable machine element is plotted in FIG. 7 as a function of time t. The curve 1c shows the time-dependence of the actual position value xist(t) of the movable machine element when using the rectangular jerk profile 1a according to FIG. 5. The curve 2c shows the actual position value of the machine element when using the cos2-shaped jerk profile 2a, and the curve 3c shows the actual position value of the machine element when using the parameter-dependent jerk profile 3a with the parameter α=0.345. As seen from the curve 3c, the actual position value xist(t) approaches the desired initial position usoll(t)=40 mm significantly faster and better than the actual position values depicted in curves 1c and 2c.

The dynamic properties of the machine can therefore be significantly improved by the parameter-dependent shape function of the invention as compared to the current technology.

FIG. 8 shows the frequency-dependent amplitudes of the Fourier transform: curve 1e was obtained with the rectangular shape function of FIG. 5, curve 2e was obtained with the associated cos2 shape function, and curve 3e was obtained with the associated parameter-dependent shape function, with α=0.356. It is evident that the frequency-dependent amplitude of the parameter-dependent shape function represents, i.a., a bandpass filter with a high damping in a frequency range around 24 Hz. The damping in this blocking region is considerably greater than with the frequency-dependent amplitude of the rectangular shape function or the cos2 shape function. Also evidently is here the significant improvement achieved by a parameter-dependent shape function as compared to the conventional rectangular or the cos2 shape functions.

It should be noted that, with the exception of the phase, a cos2 shape function is identical to a sin2 shape function. Accordingly, the cos2 function in the equations (4) and (6) can be replaced by a sin2 function.

The part of the equation (1) enclosed in square brackets describes the dependence of the transfer function A(f) of the jerk-equivalent filter R in the time intervals ta and tv depicted in FIG. 1. The process can be optimized globally by initially determining an optimized rectangular jerk profile r(t) before the machine element moves across the path defined by the controller of the machine, so that the frequencies of the bandpass filter coincide essentially with the characteristic frequencies of the machine element and/or the machine or with a subset of the characteristic frequencies. The time intervals ta and tv should be defined so as to satisfy the above condition. The already optimized rectangular jerk profile r(t) is subsequently further optimized by modification with the parameter-dependent shape function Fs(t, α).

Alternatively, the parameter-dependent shape function Fs(t, α) can also be obtained by adding a Dirac shape function F3(t) and the rectangular shape function F1(t). In principle, the entire frequency range can be covered by combining the two shape functions with a freely selected parameter α. As before, an advantageous range for the parameter α is between 0 and 1.

The Dirac shape function F3(t) is defined by the equation:

F 3 ( t ) = 1 2 δ ( t + t r 2 ) + 1 2 δ ( t - t r 2 ) ( 9 )
and visualized in FIG. 4 by the curve 4d.

The Fourier transform of the shape function F3(t) is defined by the equation (10):
ℑ(F3(t))=cos(πftr)  (10)

Accordingly, one obtains for the parameter-dependent shape function Fs(t, α):

F s ( t , α ) = α · - t r 2 t r 2 ( t t r ) + ( 1 - α ) · ( 1 2 δ ( t + t r 2 ) + 1 2 δ ( t - t r 2 ) ) ( 11 )

Moreover, one obtains for the Fourier transform of the parameter-dependent shape function Fs(t, α):

( F s ( t , α ) ) = si ( π ft r ) · α tan ( π ft r ) + ( 1 - α ) π ft r tan ( π ft r ) ( 12 )

The frequency range of a bandpass filter associated with the single parameter α can be changed through suitable selection of the parameter α. In general, the parameter-dependent shape function Fs(t, αi) can also depend on several parameters αi(i=1, 2, 3 . . . ). The frequency range of the different bandpass filters associated with an individual parameter αi can be modified by intentionally changing the corresponding parameters αi. In other words, the frequency characteristic of a particular bandpass filter can be intentionally changed by changing a corresponding parameter αi.

The analytical shape of a parameter-dependent shape function Fs(t, αi) that depends on several parameters αi can be changed by linearly superposing different basic shape functions Fi(t), such as F1(t), F2(t), and F3(t), according to the equation

F ( t , α i ) = i α i · F i ( t ) i = 1 , 2 , 3 ( 13 )
whereby as a consequence of the requirement

- t R / 2 t R / 2 r ( τ ) t = 1 ( 14 )
the parameters αi are governed by the following relationship

i α i = 1 ( 15 )

The parameters αi should be determined so as to minimize the excitations in the range of the resonance frequencies. The basic shape functions can be implemented herein as the rectangular shape function F1(t) introduced above, the cos2 shape function F2(t), and the Dirac shape function F3(t). It will be understood by those skilled in the art that a triangular shape function or a trapezoidal shape function or a similar shape function can also be used as a basic shape function.

While the invention has been illustrated and described in connection with currently preferred embodiments shown and described in detail, it is not intended to be limited to the details shown since various modifications and structural changes may be made without departing in any way from the spirit of the present invention. The embodiments were chosen and described in order to best explain the principles of the invention and practical application to thereby enable a person skilled in the art to best utilize the invention and various embodiments with various modifications as are suited to the particular use contemplated.

What is claimed as new and desired to be protected by Letters Patent is set forth in the appended claims and includes equivalents of the elements recited therein:

Ladra, Uwe, Schäfers, Elmar

Patent Priority Assignee Title
10280048, Feb 11 2015 Siemens Aktiengesellschaft Automated crane controller taking into account load- and position-dependent measurement errors
8060239, Dec 22 2005 Siemens Aktiengesellschaft Method for the determination of a rough trajectory to be followed in a positionally guided manner
8294405, Dec 13 2004 Siemens Aktiengesellschaft Method and device for guiding the movement of a moving machine element on a numerically controlled machine
8763771, Sep 30 2011 Siemens Aktiengesellschaft Active oscillation damper without direct acceleration detection
8803465, Apr 06 2012 Harris Corporation Frequency-selective command profile motion control
9740179, Sep 30 2011 Siemens Aktiengesellschaft Processing machine with vibration compensation of movable mechanical structures
Patent Priority Assignee Title
5057756, Jul 24 1988 Fanuc Ltd. Acceleration/deceleration control apparatus for servo control
5331264, Apr 15 1993 FANUC ROBOTICS NORTH AMERICA, INC Method and device for generating an input command for a motion control system
6320345, Mar 05 1998 Nikon Corporation Command trajectory for driving a stage with minimal vibration
6507165, Feb 10 2000 Fanuc Ltd Controller for machine
6825633, Jan 10 2002 Siemens Aktiengesellschaft Method for adapting jerk profiles through jerk limitation for minimizing excitations of oscillations during movement of a machine element
DE10016137,
DE10063722,
DE10118477,
DE10200680,
EP685779,
///
Executed onAssignorAssigneeConveyanceFrameReelDoc
Apr 02 2004Siemens Aktiengesellschaft(assignment on the face of the patent)
Apr 13 2004LADRA, UWESiemens AktiengesellschaftASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS 0155810810 pdf
Apr 13 2004SCHAFERS, ELMARSiemens AktiengesellschaftASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS 0155810810 pdf
Date Maintenance Fee Events
Sep 16 2010M1551: Payment of Maintenance Fee, 4th Year, Large Entity.
Dec 05 2014REM: Maintenance Fee Reminder Mailed.
Apr 24 2015EXP: Patent Expired for Failure to Pay Maintenance Fees.


Date Maintenance Schedule
Apr 24 20104 years fee payment window open
Oct 24 20106 months grace period start (w surcharge)
Apr 24 2011patent expiry (for year 4)
Apr 24 20132 years to revive unintentionally abandoned end. (for year 4)
Apr 24 20148 years fee payment window open
Oct 24 20146 months grace period start (w surcharge)
Apr 24 2015patent expiry (for year 8)
Apr 24 20172 years to revive unintentionally abandoned end. (for year 8)
Apr 24 201812 years fee payment window open
Oct 24 20186 months grace period start (w surcharge)
Apr 24 2019patent expiry (for year 12)
Apr 24 20212 years to revive unintentionally abandoned end. (for year 12)